In phylogenetics and phylogenomics, a common assumption made during biomolecular sequence analysis is that a single “global” evolutionary model sufficiently describes the evolution of all sites within a given dataset. However, numerous evolutionary processes can violate this assumption and introduce local model mis-specification. While both traditional statistical methods and machine learning approaches have been employed to address this issue, they are typically constrained by specific modeling assumptions and often are specialized for one or a few closely related tasks, limiting their applicability. In this study, we introduce REVEAL (“REsampling and Visual EvALuation”), a general-purpose statistical framework for detecting and mapping local model mis-specification during biomolecular sequence analysis. REVEAL does not impose additional modeling assumptions beyond those used during global model-based sequence analysis. REVEAL leverages sequence-aware statistical resampling techniques to extract a local support matrix along the input sequences, enabling the identification of potential local model violations. Performance benchmarking using simulation experiments demonstrates that REVEAL achieves robust type I and type II error, with as high as 90% precision and 85% recall across a range of experimental conditions that have different sources of local model mis-specification. We also employ REVEAL to analyze genomic sequence data for mouse and mosquito, and REVEAL detects local model violations that align with findings from previously published studies.